{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ILBQLXKXZEMOKKKKK5A44Q6NCK","short_pith_number":"pith:ILBQLXKX","schema_version":"1.0","canonical_sha256":"42c305dd57c918e5294a5741ce43cd12b8a7b789044010d357baf6bac95fe44a","source":{"kind":"arxiv","id":"2608.05886","version":1},"attestation_state":"computed","paper":{"title":"CodeGrep: An RL-Trained Retrieval Agent for LLM Coding Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Wuya Chen, Yang Cao, Yihao Yang, Yue Lin","submitted_at":"2026-08-06T11:07:42Z","abstract_excerpt":"Modern LLM coding agents such as Claude Code and OpenHands share a common inefficiency: they spend much of their token budget finding the file to patch, rather than patching it. On SWE-Bench Verified, a 30B OpenHands agent averages 23 rounds and 631K tokens per resolved issue, with many calls spent on grep, glob, and view_file during repository exploration. We introduce CodeGrep, a 14B retrieval agent trained end-to-end with GRPO to issue multi-turn parallel grep, glob, and read tool calls and return candidate files to a frozen downstream coding agent.\n  On all 500 SWE-Bench Verified instances"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2608.05886","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SE","submitted_at":"2026-08-06T11:07:42Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"961343486ca13e77262473eb1c05aa4cf63d1ac297d29fee1ef1097f3ea31a97","abstract_canon_sha256":"248552bf064de0968fb2438f8964cc5a9c4c7af65e28583fc9a6c80319e15232"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-07T00:53:12.296664Z","signature_b64":"Okh8oyMP/bdKJ6DNiwVJbZ3GoxcSMvwMCKiF9wLJVsuqlwfv2BuGzTn8c4qQgtkntGhCLwBo5dyMeReXiVCNDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"42c305dd57c918e5294a5741ce43cd12b8a7b789044010d357baf6bac95fe44a","last_reissued_at":"2026-08-07T00:53:12.295196Z","signature_status":"signed_v1","first_computed_at":"2026-08-07T00:53:12.295196Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CodeGrep: An RL-Trained Retrieval Agent for LLM Coding Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Wuya Chen, Yang Cao, Yihao Yang, Yue Lin","submitted_at":"2026-08-06T11:07:42Z","abstract_excerpt":"Modern LLM coding agents such as Claude Code and OpenHands share a common inefficiency: they spend much of their token budget finding the file to patch, rather than patching it. On SWE-Bench Verified, a 30B OpenHands agent averages 23 rounds and 631K tokens per resolved issue, with many calls spent on grep, glob, and view_file during repository exploration. We introduce CodeGrep, a 14B retrieval agent trained end-to-end with GRPO to issue multi-turn parallel grep, glob, and read tool calls and return candidate files to a frozen downstream coding agent.\n  On all 500 SWE-Bench Verified instances"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.05886","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2608.05886/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2608.05886","created_at":"2026-08-07T00:53:12.296739+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.05886v1","created_at":"2026-08-07T00:53:12.296739+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.05886","created_at":"2026-08-07T00:53:12.296739+00:00"},{"alias_kind":"pith_short_12","alias_value":"ILBQLXKXZEMO","created_at":"2026-08-07T00:53:12.296739+00:00"},{"alias_kind":"pith_short_16","alias_value":"ILBQLXKXZEMOKKKK","created_at":"2026-08-07T00:53:12.296739+00:00"},{"alias_kind":"pith_short_8","alias_value":"ILBQLXKX","created_at":"2026-08-07T00:53:12.296739+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK","json":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK.json","graph_json":"https://pith.science/api/pith-number/ILBQLXKXZEMOKKKKK5A44Q6NCK/graph.json","events_json":"https://pith.science/api/pith-number/ILBQLXKXZEMOKKKKK5A44Q6NCK/events.json","paper":"https://pith.science/paper/ILBQLXKX"},"agent_actions":{"view_html":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK","download_json":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK.json","view_paper":"https://pith.science/paper/ILBQLXKX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.05886&json=true","fetch_graph":"https://pith.science/api/pith-number/ILBQLXKXZEMOKKKKK5A44Q6NCK/graph.json","fetch_events":"https://pith.science/api/pith-number/ILBQLXKXZEMOKKKKK5A44Q6NCK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK/action/storage_attestation","attest_author":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK/action/author_attestation","sign_citation":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK/action/citation_signature","submit_replication":"https://pith.science/pith/ILBQLXKXZEMOKKKKK5A44Q6NCK/action/replication_record"}},"created_at":"2026-08-07T00:53:12.296739+00:00","updated_at":"2026-08-07T00:53:12.296739+00:00"}